Engineering Papers⌕ Search

SEARCH · Engineering Papers

Results for “adaptive learning”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,261 records · Page 70

Denoising diffusion models with geometry adaptation for high fidelity calorimeter simulation

Simulation is crucial for all aspects of collider data analysis, but the available computing budget in the High Luminosity LHC era will be severely constrained. Generative machine learning models may act as surrogates to replace physics-based full simulation of particle detectors, and diffusion models have recently emerged as the state of the art for other generative tasks. Here, we introduce CaloDiffusion, a denoising diffusion model trained on the public CaloChallenge datasets to generate calorimeter showers. Our algorithm employs 3D cylindrical convolutions, which take advantage of symmetries of the underlying data representation. To handle irregular detector geometries, we augment the diffusion model with a new geometry latent mapping (GLaM) layer to learn forward and reverse transformations to a regular geometry that is suitable for cylindrical convolutions. The showers generated by our approach are nearly indistinguishable from the full simulation, as measured by several different metrics.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

GL4U: Using Space Omics Data to Provide Bioinformatics Training for Students and Educators

NASA’s GeneLab project provides researchers open access to space-relevant experiment multi-omics data that can be mined to understand the effects of spaceflight on biological systems. To maximize the number of scientists who understand and utilize GeneLab data and data processing pipelines, GeneLab has created GeneLab for Colleges and Universities (GL4U). GL4U provides space biology-relevant training in bioinformatics to the next generation of scientists through direct and indirect approaches. The GeneLab team plans to host two annual data processing bootcamps, one for college-level students (direct) and one for college educators (indirect – training of trainers), in which participants learn to analyze GeneLab’s space-relevant omics data. The GL4U direct training pilot program was conducted in June 2021. During the pilot, students participated in a week-long bootcamp consisting of space biology-specific lectures and hands-on instruction using Jupyter Notebooks to analyze RNA sequence data. This pilot demonstrated the capacity of GL4U for training young scientists and encouraging data re-use. In June 2022, GL4U partnered with Jet Propulsion Laboratory’s (JPL) Planetary Protection Center of Excellence to conduct the indirect training pilot program by training educators at historically black colleges and universities (HBCUs) and minority serving institutions (MSIs). During the educator pilot, participants received materials, training, and will be provided the necessary compute resources to enable them to run the bootcamp at their home institutions or alternatively to adapt the content to implement within existing courses, thereby extending the reach of this initiative. The GL4U training program provides undergraduate students from underrepresented groups the opportunity to learn about NASA and Space Biology, and to enhance their career prospects by gaining hands-on experience analyzing omics data, a skillset that is highly applicable and marketable in the life sciences. Pre- and post-bootcamp surveys were completed by all participants and show the overwhelming success of the bootcamps.

Amanda M. Saravia-Butler↗

Standard Component Commodity Usage Guides (CUGs) for Charge-Coupled Devices (CCDs)

Communicating risk and reliability information within NASA Goddard’s missions is crucial to continuing the development of new technology and ensuring GSFC missions achieve their science objectives. The Reliability and Risk Assessment Branch (Code 371) within the SMA directorate is one of GSFC’s leaders in communicating this information to missions, and one tool 371 engineers use to provide this information is the Commodity Usage Guides (CUGs). CUGs are helpful in compiling various critical parameters and heritage information for standard payload components, ensuring project personnel inherit the lessons learned from previous experiences. The purpose of this project was to create CUGs for CCDs used within GSFC missions. To formulate a CUG applicable data was aggregated into the CUG template for Charge-Coupled Devices (CCDs)--electro-optical components used to convert photons into digital images and data. CCDs can easily be adapted to a variety of mission requirements making them widely used at GSFC, and therefore suitable for CUGs. Once the CUGs for CCDs are completed the information can be distributed to engineers, scientists and other project members to ensure the lessons learned are passed on to the next mission.

Christman, Jerika H.↗

Countermeasures to Enhance Sensorimotor Adaptability

During exploration-class missions, sensorimotor disturbances may lead to disruption in the ability to ambulate and perform functional tasks during the initial introduction to a novel gravitational environment following a landing on a planetary surface. The goal of our current project is to develop a sensorimotor adaptability (SA) training program to facilitate rapid adaptation to novel gravitational environments. We have developed a unique training system comprised of a treadmill placed on a motion-base facing a virtual visual scene that provides an unstable walking surface combined with incongruent visual flow designed to enhance sensorimotor adaptability. We have conducted a series of studies that have shown: Training using a combination of modified visual flow and support surface motion during treadmill walking enhances locomotor adaptability to a novel sensorimotor environment. Trained individuals become more proficient at performing multiple competing tasks while walking during adaptation to novel discordant sensorimotor conditions. Trained subjects can retain their increased level of adaptability over a six months period. SA training is effective in producing increased adaptability in a more complex over-ground ambulatory task on an obstacle course. This confirms that for a complex task like walking, treadmill training contains enough of the critical features of overground walking to be an effective training modality. The structure of individual training sessions can be optimized to promote fast/strategic motor learning. Training sessions that each contain short-duration exposures to multiple perturbation stimuli allows subjects to acquire a greater ability to rapidly reorganize appropriate response strategies when encountering a novel sensory environment. Individual sensory biases (i.e. increased visual dependency) can predict adaptive responses to novel sensory environments suggesting that customized training prescriptions can be developed to enhance adaptability. These results indicate that SA training techniques can be added to existing treadmill exercise equipment and procedures to produce a single integrated countermeasure system to improve performance of astro/cosmonauts during prolonged exploratory space missions.

Bloomberg, J. J.↗

Dynamic Decarbonization through Autonomous Physics-Centric Deep Learning and Optimization of Building Operations (Abstract only)

This project directly addresses the primary goal of Area of Interest 2 in the CRADA call: to advance optimization-based integrated energy management systems in commercial and residential buildings. Pacific Northwest National Laboratory (PNNL) and its industry partner PassiveLogic aim to accomplish this by reaching three key objectives. First, to ensure a broad impact in the building controls industry, PNNL will extend its open-source library for predictive control synthesis by augmenting its capabilities with data-driven self-learning of building models and auto-calibration of predictive controllers. The effort will focus on building use cases selected in collaboration with PassiveLogic. The team will specifically address the development of methods for data-driven adaptation of building models, investigation of model architectures that best address specific building types, and automated synthesis of differentiable predictive controllers that optimize diverse objectives. Second, PNNL will collaborate with PassiveLogic to integrate the aforementioned methods with PasiveLogic’s advanced controls platform. The collaborative integration effort will inform the developments under the first objective by providing specific data on the attainable performance of model learning on resource-constrained edge computing platforms. This software integration effort will increase the technical maturity of the developed libraries by exploring the use of software integration tools and methods. Third, PNNL and PassiveLogic will work to improve the technology readiness of the developed predictive controllers by testing their performance in relevant test environments, such as high-fidelity simulation, hardware in the loop, and actual test buildings.

32 ENERGY CONSERVATION, CONSUMPTION, AND UTILIZATI↗

Sequential Decision Making (SDM) for Mesh Refinement and Model Selection in Multiscale, Multi-Physics Applications

Intelligent automation and decision support are needed to enhance computational efficiency and robustness in multiscale and multi-physics problems, including materials science, manufacturing, and climate and weather modeling. Current scientific computing approaches for enabling decisions by scientists fail to explore the role of learning, reasoning, and probabilistic planning. Often these decisions are not performed in real-time during the computation but are made prior to the start of the computation, which must be interrupted in order to make changes to the prior choices. Such interruptions at different stages of the computation increase the total computing time and the need for a human expert to frequently monitor the results. State of art scientific computing methods consist of rule-based algorithms that cannot automatically adapt to a dynamically changing computing environment. The development of a Sequential Decision Making (SDM) framework will automate scientific computing by optimizing the policies for mesh refinement, time-stepping, model and algorithm selection, resource allocation, and pre and post-processing. Our agent SDM framework for scientific computing will consist of data-driven learning (Classifier), automated reasoning (contextual knowledge), and probabilistic planning (Reinforcement Learning). In this project, we focused on three problems to demonstrate our SDM framework on a set of ordinary and partial differential equations. Classification of Lorenz system regions using Feed-Forward Neural Networks examined learning in the SDM framework. On the other hand, reasoning and planning in the SDM framework were used in two problems: adaptive time-stepping for nonlinear ODEs using on-policy RL algorithms, and adaptive mesh refinement for 2-D PDEs using off-policy RL algorithms.

97 MATHEMATICS AND COMPUTING↗

Tackling Climate Change with Machine Learning

Climate change is one of the greatest challenges facing humanity, and we, as machine learning (ML) experts, may wonder how we can help. Here we describe how ML can be a powerful tool in reducing greenhouse gas emissions and helping society adapt to a changing climate. From smart grids to disaster management, we identify high impact problems where existing gaps can be filled by ML, in collaboration with other fields. Our recommendations encompass exciting research questions as well as promising business opportunities. We call on the ML community to join the global effort against climate change.

54 ENVIRONMENTAL SCIENCES↗

Adapting Dismantling and Decommissioning Strategies to a Variety of Nuclear Fuel Cycle Facilities - 12237

AREVA has accumulated over 20 years of experience in managing and operating fuel cycle facilities Decontamination and Decommissioning (D and D) projects of many different types and a variety of scales, both as facility owner (at La Hague for example) and as prime contractor to external customers such as the French Atomic Energy Commission (at Marcoule). A specific Business Unit was created in 2008 to capitalize on this experience and to concentrate - in one division - the specific skills required to be successful and cost effective in decommissioning projects. Indeed one of the key lessons learned in the past decades is that decommissioning is a significantly different business as compared to normal operations of a nuclear facility. Almost all the functions of a project need to be viewed from a different angle, challenged and adapted consequently in order to optimize costs and schedule. Three examples follow to illustrate the point: Safety management needs to take into account the ever changing configuration of a plant under D and D (a quite new situation for the authorities). Production of waste is significantly different in term of volume, activities, conditioning and disposal path. Technology is important but technical issues are often less critical than good management and planning. Further examples and lessons learned are developed through reviewing the projects experience basis. AREVA has a long and vast experience in the cleanup and dismantling of a number of very large and complex nuclear facilities. This effort focused initially on AREVA's own plants and is expanding now to other customers. The setup of a specific Business Unit in 2008 to takeover this business allowed concentration of the skills and the lessons learned in a dedicated division so as to provide the best means to optimize safety, performance, costs and schedules. Indeed transitioning from operations to D and D of a nuclear facility is a quantum leap. The assistance from specialized teams can bring significant cost savings. (authors)

11 NUCLEAR FUEL CYCLE AND FUEL MATERIALS↗

LIMITING ACCRETION ONTO MASSIVE STARS BY FRAGMENTATION-INDUCED STARVATION

Massive stars influence their surroundings through radiation, winds, and supernova explosions far out of proportion to their small numbers. However, the physical processes that initiate and govern the birth of massive stars remain poorly understood. Two widely discussed models are monolithic collapse of molecular cloud cores and competitive accretion. To learn more about massive star formation, we perform and analyze simulations of the collapse of rotating, massive, cloud cores including radiative heating by both non-ionizing and ionizing radiation using the FLASH adaptive-mesh refinement code. These simulations show fragmentation from gravitational instability in the enormously dense accretion flows required to build up massive stars. Secondary stars form rapidly in these flows and accrete mass that would have otherwise been consumed by the massive star in the center, in a process that we term fragmentation-induced starvation. This explains why massive stars are usually found as members of high-order stellar systems that themselves belong to large clusters containing stars of all masses. The radiative heating does not prevent fragmentation, but does lead to a higher Jeans mass, resulting in fewer and more massive stars than would form without the heating. This mechanism reproduces the observed relation between the total stellar mass in the cluster and the mass of the largest star. It predicts strong clumping and filamentary structure in the center of collapsing cores, as has recently been observed. We speculate that a similar mechanism will act during primordial star formation.

79 ASTRONOMY AND ASTROPHYSICS↗

Dynamic Model Development of a Wind Power Plant Using Neural Net Method to Forecast Wind Power Output (CRADA Final Report)

This project is intended to model wind power plant based on monitored data at the wind power plant. This project will promote the university research in Renewable Energy area and trains the future highly qualified engineers. The dynamic model will be based on neural net model with the input from the two met towers (12 inputs), and the number of turbines in operation (one input). The overall input will be 13 inputs to drive the simulations. The output power at the point of interconnection will be used to tune the neural net weight coefficients. Two neural net concepts will be investigated (the back propagation neural net and the dynamic recurrent neural net with feedback).

17 WIND ENERGY↗

Visual orientation performances of desert ants (Cataglyphis bicolor) toward astromenotactic directions and horizon landmarks

Experimental data, on the visual orientation of desert ants toward astromenotactic courses and horizon landmarks involving the cooperation of different direction finding systems, are given. Attempts were made to: (1) determine if the ants choose a compromise direction between astromenotactic angles and the direction toward horizon landmarks when both angles compete with each other or whether they decide alternatively; (2) analyze adaptations of the visual system to the special demands of direction finding by astromenotactic orientation or pattern recognition; and (3) determine parameters of visual learning behavior. Results show separate orientation mechanisms are responsible for the orientation of the ant toward astromenotactic angles and horizon landmarks. If both systems compete with each other, the ants switch over from one system to the other and do not perform a compromise direction.

Wehner, R.↗

Model Based Systems Engineering on the Europa Mission Concept Study

At the start of 2011, the proposed Jupiter Europa Orbiter (JEO) mission was staffing up in expectation of becoming an official project later in the year for a launch in 2020. A unique aspect of the pre-project work was a strong emphasis and investment on the foundations of Model-Based Systems Engineering (MBSE). As so often happens in this business, plans changed: NASA's budget and science priorities were released and together fundamentally changed the course of JEO. As a result, it returned to being a study task whose objective is to propose more affordable ways to accomplish the science. As part of this transition, the question arose as to whether it could continue to afford the investment in MBSE. In short, the MBSE infusion has survived and is providing clear value to the study effort. By leveraging the existing infrastructure and a modest additional investment, striking advances in the capture and analysis of designs using MBSE were achieved. In the process, the need to remain relevant in the new environment has brought about a wave of innovation and progress. The effort has reaffirmed the importance of architecting. It has successfully harnessed the synergistic relationship of architecting to system modeling. We have found that MBSE can provide greater agility than traditional methods. We have also found that a diverse 'ecosystem' of modeling tools and languages (SysML, Mathematica, even Excel) is not only viable, but an important enabler of agility and adaptability. This paper will describe the successful application of MBSE in the dynamic environment of early mission formulation, the significant results produced and lessons learned in the process.

Bayer, Todd J.↗

A Dynamic Reduction Network for Point Clouds

Classifying whole images is a classic problem in machine learning, and graph neural networks are a powerful methodology to learn highly irregular geometries. It is often the case that certain parts of a point cloud are more important than others when determining overall classification. On graph structures this started by pooling information at the end of convolutional filters, and has evolved to a variety of staged pooling techniques on static graphs. In this paper, a dynamic graph formulation of pooling is introduced that removes the need for predetermined graph structure. It achieves this by dynamically learning the most important relationships between data via an intermediate clustering. The network architecture yields interesting results considering representation size and efficiency. It also adapts easily to a large number of tasks from image classification to energy regression in high energy particle physics.

72 PHYSICS OF ELEMENTARY PARTICLES AND FIELDS↗

French management model for sustainable transportation of used fuel - 15376

Safe and efficient transportation of radioactive material is key to the success of our industry. Used fuel transports are inherently complex due to the nature of the material and the high visibility. Only a few countries have developed comprehensive used fuel transportation systems that require a wide range of equipment, services and expertise. In the many countries, such as the Unites States, where interim consolidated storage, recycling or geological repository are not available yet, similar systems will have to be implemented. For more than 50 years AREVA TN has safely shipped more than 7,000 used fuel transport casks. The transportation system that was initially developed in the 1970's has been adapted and enhanced over the years to meet more restrictive regulatory requirements and evolving customer needs, and to address public concerns. The numerous 'lessons learned' have offered data and guidance that have allowed for also efficient and consistent improvement over the decades. Based on AREVA TN extensive expertise, this paper will describe the different phases and milestones that need to be met to set up, license and operate a successful used fuel transportation management system. Transportation of spent fuel in France could not be performed effectively and efficiently without strong collaboration with Electricite de France (EDF) who plays a critical role as the shipper. EDF will provide a nuclear operator's perspective on the most critical aspects of a successful used fuel transportation program. (authors)

12 MANAGEMENT OF RADIOACTIVE AND NON-RADIOACTIVE W↗

Towards in-situ certification of additively manufactured parts: the vital roles of physics-based and data-driven models

Certifying additively manufactured (AM) parts in-situ at the completion of a build is an enticing prospect, as it can help reduce the high costs associated with post-build testing and evaluation. However, achieving this goal presents significant challenges that may keep it aspirational for the foreseeable future. Nonetheless, incremental progress can pave the way forward. A critical aspect of in-situ certification involves continuous quality checking due to the random nature of the AM process and the difficulties in detecting defects or anomalies once layers are built over. While real-time in-situ monitoring strategies assisted by machine learning (ML) play a pivotal role in auditing part quality, they must ideally be supported by real-time (or near real-time) adaptive process control enabled by ML-assisted decision-making. By analyzing in-situ monitoring data in real-time (or near real- time) to dynamically adjust manufacturing parameters, such intervention can ensure AM parts are built to meet stringent certification standards. This can be achieved virtually by using high-fidelity performance models for the physical testing and evaluation tasks. In this short editorial, we discuss the key contributions made by data-driven and physics-based models in providing intelligence to the monitoring and process control tasks underpinning in-situ certification and in the simulation of the build’s performance under test and service conditions. While our focus lies in metal AM, the concepts discussed here are also relevant to other AM processes.

: In-situ monitoring↗

Deep-Learning-derived Boundary Layer Height from Meteorological Data over the SGP, GOAMAZON, CACTI

The planetary boundary-layer (PBL) height (PBLH) is an important parameter for various meteorological and climate studies. This study presents a multi-structure deep neural network (DNN) model, designed to estimate PBLH by integrating morning temperature profiles with surface meteorological observations. The DNN model is developed by leveraging a rich data set of PBLH derived from long-standing radiosonde records and augmented with high-resolution micropulse lidar and Doppler lidar observations. We access the performance of the DNN with an ensemble of 10 members, each featuring distinct hidden layer structures, which collectively yield a robust 27-year PBLH data set over the Southern Great Plains from 1994 to 2020. The influence of various meteorological factors on PBLH is rigorously analyzed through the importance test. Moreover, the DNN model's accuracy is evaluated against radiosonde observations and juxtaposed with conventional remote-sensing methodologies, including Doppler lidar, ceilometer, Raman lidar, and micropulse lidar. The DNN model exhibits reliable performance across diverse conditions and demonstrates lower biases relative to remote-sensing methods. In addition, the DNN model, originally trained over a plain region, demonstrates remarkable adaptability when applied to the heterogeneous terrains and climates encountered during the GoAmazon (tropical rainforest) and CACTI (middle-latitude mountain) campaigns. These findings demonstrate the effectiveness of deep learning models in estimating PBLH, enhancing our understanding of boundary-layer dynamics with implications for enhancing the representation of PBL in weather forecasting and climate modeling.

54 ENVIRONMENTAL SCIENCES↗

Optimization of carbon capture systems using surrogate models of simulated processes.

With increasing demand placed on power generation plants to reduce carbon dioxide (CO2) emissions, processes to separate and capture CO2 for eventual sequestration are highly sought after. Carbon capture processes impart a parasitic load on the power plants; it is estimated that this would increase the cost of electricity from existing pulverized coal plants anywhere from 71-85 percent [1]. The National Energy and Technology Lab (NETL) is working to lower this to below a 30 percent increase. To reach this goal, work is being done not only to accurately simulate these processes, but also to leverage those accurate and detailed simulations to design optimal carbon capture processes. The major challenges include the lack of accurate algebraic models of the processes, computationally costly simulations, and insufficiently robust simulations. The first challenge bars the use of provable derivative-based optimization algorithms. The latter two can either lead to difficult or impossible direct derivative-free optimization. To overcome these difficulties, we take a more indirect method to solving this problem by, first, generating an accurate set of algebraic surrogate models from the simulation then using derivative-based solvers to optimize the surrogate models. We developed a method that uses derivative-based and derivative-free optimization alongside machine learning and statistical techniques to generate the set of low-complexity surrogate models using data sampled from detailed simulations. The models are validated and improved through the use of derivative-free solvers to adaptively sample new simulation points. The resulting surrogate models can then be used in a superstructure-based process synthesis and solved using derivative-based methods to optimize carbon capture processes.

01 COAL, LIGNITE, AND PEAT↗

On Transfer Learning Of Neural Networks Using Bi-Fidelity Data For Uncertainty Propagation

Due to their high degree of expressiveness, neural networks have recently been used as surrogate models for mapping inputs of an engineering system to outputs of interest. Once trained, neural networks are computationally inexpensive to evaluate and remove the need for repeated evaluations of computationally expensive models in uncertainty quantification applications. However, given the highly parameterized construction of neural networks, especially deep neural networks, accurate training often requires large amounts of simulation data that may not be available in the case of computationally expensive systems. In this paper, to alleviate this issue for uncertainty propagation, we explore the application of transfer learning techniques using training data generated from both high- and low-fidelity models. We explore two strategies for coupling these two datasets during the training procedure, namely, the standard transfer learning and the bi-fidelity-weighted learning. In the former approach, a neural network model mapping the inputs to the outputs of interest is trained based on the low-fidelity data. The high-fidelity data are then used to adapt the parameters of the upper layer(s) of the low-fidelity network, or train a simpler neural network to map the output of the low-fidelity network to that of the high-fidelity model. In the latter approach, the entire low-fidelity network parameters are updated using data generated via a Gaussian process model trained with a small high-fidelity dataset. The parameter updates are performed via a variant of stochastic gradient descent with learning rates given by the Gaussian process model. Using three numerical examples, we illustrate the utility of these bi-fidelity transfer learning methods where we focus on accuracy improvement achieved by transfer learning over standard training approaches.

97 MATHEMATICS AND COMPUTING↗